详细内容
iroMind
Manrope
Open Deep Research
Build Together. Evolve Forever.
We don't deliver AI. We build it with you.
We're building open, honest, and patient Agentic AI together with developers — not above them. This is day one.
DeepResearch is live - here's what we built\
\
Try Now
MiroThinker
MiroThinker is a series of open source SoTA Deep Research models. It surpasses all open source models with a 60.2% Avg@8 score for the challenging GAIA benchmark.
MiroTrain/MiroRL
MiroTrain and MiroRL are the training infrastructure for our MiroThinker. They enable users to develop long context SFT/RL models up to 64K context out of the box.
MiroFlow
Miroflow is a SoTA agent framework that powers both propriety models and Mirothinker for tool-use agent tasks. It features a fair and reproducible GAIA score of 82.4%.
MiroVerse
MiroVerse is a collection of high-quality verified solution traces for agent tasks we collected for developing MiroThinker. It encompass an ever-growing list of tasks from subdomains including STEM subjects, reasoning and hard fact seesking.
\
\
Serial Entrepreneur Tianqiao Chen Presents Vision for “Discoverative” Intelligence at AIAS2025\
\
Oct 31, 2025\
\
Learn More
\
\
MiroMind’s Open Deep Research Framework Tops FutureX Benchmark, Heralds a New Paradigm in AI\
\
Sep 19, 2025\
\
Learn More
\
\
MiroFlow v0.2\
\
Aug 25, 2025\
\
Learn More
\
\
MiroMind Research Agent v0.1\
\
Aug 7, 2025\
\
Learn More
Research
Community

Scan QR code
Legal
© Copyright 2025 All Rights Reserved
Cookie Settings
We use cookies to personalize content, run ads, and analyze traffic. Read our Cookie Policy.
